Among first 14, 9are correct
Amonng first 14, first 2 are correct
Among first 14,8 are correct
Among first 14, 5 are correct

actually when we search for data science, we should get all indices from 0 to 9, but that is not the case here and only some results are correct in all the four models

Among first 14,10 are correct ,previously 9-->increase in accuracy

here first 7 results are correct, among first 11, 9 correct and among first 13, 10 are correct this is littile bit better than simple semantic model but slow

Among first 14, 10 are correct, in previous simple semantic model 2 are correct--->increase in accuracy

this process is very slow, but helps to deal at minute level

you can try this method with model2 and model3 also, there is definitly increase in accuracy compared to simple semantic models 2 and 3

using faiss

https://www.youtube.com/watch?v=sKyvsdEv6rk&t=1230s #see the video to understand faiss

here we have used model0 for sentence embedding but if we use any model among above 4 we would not get accurate results,faisss is about dealing with large number of vectors and fastness, not accuracy, so we proceed further

so now we are going to combine semantic model for semantic searching and faiss for dealing large number of vectors and fastness and temp3(ngrams) for accuracy

so it is fast compared to semantic with temp3 method and accurate than normal semantic and normal faiss
The below one is rough work